Antibody-based therapies for cancer have the potential of higher specificity and lower side effect profile as compared to conventional drugs. The reason is a precise distinction between normal and neoplastic cells by antibodies and the fact, that their mode of action relies on less toxic immunological anti-tumor mechanisms, such as complement activation and recruitment of cytotoxic immune cells.
Targets for antibody-based therapies need to have particular qualities, which form the basis for proper discrimination between normal and neoplastic cells. Obviously, a target with either exclusive restriction to tumor cells and entirely undetectable on normal tissues is ideal for the development of efficient and safe antibody therapeutics. In another aspect, a high-level overexpression may be the basis for the therapeutic window and low side effects exemplified by the human epidermal growth factor receptor type 2 (HER-2), which as a result of gene amplification is a good target for the antibody trastuzumab (Herceptin).
Other targets for antibodies which are either already approved or in clinical development for tumor therapy have distinct qualities, which are not based on a numeric overexpression of target molecules on tumor cells. In the case of antibodies to the proteoglycan MUC-1, a peptide repeat epitope in the backbone of the target is underglycosylated in tumor cells and thus altered to its normal counterpart. In the case of antibodies to CD20 (rituximab), CD52 (Campath-1H) and CD22 (epratuzumab), antibody targets have comparable expression levels on tumor cells and normal lymphocytes. Here, the ablation of normal cells by the antibody is tolerable since target-negative stem cells restore the normal lymphocyte repertoire. Other examples of differential accessibility of antibody targets are carcinoembryonal antigen (CEA) and carboanhydrase IX (CA9). Both antigens are expressed on normal epithelia of colon and kidney, respectively. However, radioactively labeled imaging antibodies do distinguish well between tumor and normal tissue, and cytotoxic antibodies are well tolerated. This is most likely due to a restricted expression of CA9 and CEA on the luminal side of normal epithelial tissue where IgG antibodies do not have access. Also antigen epithelial cell adhesion molecule (Ep-CAM) belongs to this category. As a homotypic cell adhesion molecule for epithelial cells it is localized in the intercellular space. Intriguingly, whereas high-affinity anti-Ep-CAM antibodies are very toxic, intermediate-affinity antibodies are well tolerated. This suggests accessibility of the Ep-CAM target on normal cells but also indicates that kinetics of antibody binding may open a therapeutic window.
One possibility is that other epithelial cell-specific proteins involved in cell/cell adhesion may be also attractive for antibody approaches, since they may be barely accessible in well-structured epithelia to antibodies but become exposed on tumor cells. We therefore analyzed proteins involved in organizing epithelial tissue architecture for their suitability as targets for therapeutic antibodies. A protein, which particularly attracted our attention is claudin 18.
The claudin 18 (CLD18) molecule (Genbank accession number: splice variant 1 (CLD18A1): NP_057453, NM_016369, and splice variant 2 (CLD18A2): NM_001002026, NP_001002026) is an integral transmembrane protein with a molecular weight of approximately 27.9/27.72 kD. Claudins are integral membrane proteins located within the tight junctions of epithelia and endothelia. Tight junctions organize a network of interconnected strands of intramembranous particles between adjacent cells. In tight junctions, occludin and claudins are the most prominent transmembrane protein components. Due to their strong intercellular adhesion properties they create a primary barrier to prevent and control the paracellular transport of solutes and restrict the lateral diffusion of membrane lipids and proteins to maintain cellular polarity. Tight junction forming proteins are critically involved in organizing epithelial tissue architecture. We assumed that such proteins may be barely accessible to antibodies in well-structured epithelia but become exposed on tumor cells.
CLD18 is a tetraspanin and has as such 4 hydrophobic regions. We have generated data indicating that CLD18 displays several different conformations, which may be selectively addressed by antibodies. One conformation (CLD18-Conformation-1) implies, that all four hydrophobic regions serve as regular transmembrane domains (TM) and two extracellular loops (loop1 embraced by hydrophobic region 1 and hydrophobic region 2; loop2 embraced by hydrophobic regions 3 and 4) are formed, as described for the vast majority of claudin family members. A second conformation (CLD18-Conformation-2) implies that, as described for PMP22, another member of the tetraspanin family (Taylor et al., J. Neurosc. Res. 62:15-27, 2000), that the second and third hydrophobic domains do not fully cross the plasma membrane so that portion (loopD3) in between the first and fourth transmembrane domain is extracellular. A third conformation (CLD18-Conformation-3) implies, a large extracellular domain with two internal hydrophobic regions embraced by the first and fourth hydrophobic region, which serve as regular transmembrane domains. Due to the presence of classical N-glycosylation site in loopD3 the Claudin-18 topology variants CLD18 topology-2- and CLD18 topology-3 harbour an additional extracellular N-glycosylation site.
Another level of complexity is added to CLD18 molecule by the presence of two different splice variants, which are described in mouse and in human (Niimi, Mol. Cell. Biol. 21:7380-90, 2001). The splice variants CLD18A1 and CLD18A2 differ in the first 21 N-terminal amino acids, which comprise the first TM and loop1, whereas the primary protein sequence of the C-terminus is identical.
CLD18A1 is selectively expressed on normal lung and stomach epithelia, whereas CLD18A2 is expressed only on gastric cells (Niimi, Mol. Cell. Biol. 21:7380-90, 2001). Most importantly, CLD18A2 is restricted to the differentiated short-lived cells of stomach epithelium but is devoid from the gastric stem cell region. Using sensitive RT-PCR, we have shown that both variants are not detectable at all in any other normal human organ, but are robustly expressed in several cancer types including stomach, esophageal, pancreatic and lung tumors as well as human cancer cell lines. Expression is most prominent in the adenocarcinoma subtypes of these indications.
The molecular weight of the protein differs in some cancers and adjacent normal tissue. The higher molecular weight protein observed in healthy tissue can be transferred into the same molecular weight as observed in cancer by treating tissue lysates with the deglycosylating compound PNGase F. This suggests, that CLD18 is less N-glycosylated in cancer as compared to its normal tissue counterpart. This structural difference is likely to give rise to an altered epitope. A classical N-glycosylation motif is in position aa 116 within the loopD3 domain of the molecule.
The terms “CLD18” and “CLD18-variant” according to the invention shall encompass (i) CLD18-splice variants, (ii) CLD18-N-glycosylation variants, (iii) CLD18-conformation variants, (iv) CLD18-free and homotypically/heterotypically associated variants localized at intercellular tight junctions and (v) CLD18-cancer related and CLD18-non-cancer cell related variants.
The molecular and functional characteristics of CLD18 make this molecule a highly interesting target for antibody based cancer therapy. These are in particular (i) the absence of CLD18 from the vast majority of toxicity relevant normal tissues, (ii) the restriction of CLD18A2 variant expression to a dispensible cell population as differentiated gastric cells, which can be replenished by target-negative stem cells of the stomach, (iii) hints to potential differential glycosylation between normal and neoplastic cells, and (iv) the presence of different conformational topologies. Moreover, the role of CLD18 as tight junction protein may further contribute to a good therapeutic window. Because tumor cells express claudins but often do not form classical tight junctions by homotypic and heterotypic association of claudins as found in normal epithelial tissue, tumor cells may have a considerable pool of free claudin that is amenable to extracellular antibody binding and immunotherapy. It is possible that binding epitopes of claudins in healthy epithelium are shielded within tight junctions from the access by such antibodies.
The object of the invention is to provide antibodies useful for therapy of diseases wherein CLD18 is expressed, such as tumor diseases. The antibodies described herein have also utility in diagnosing such diseases.